August 2025 — page 2 of 2
- New detection method protects vision-language models from jailbreak attacks without training data· AI Security
- Researchers Discover Method to Bypass Safety Controls in AI Language Models Through Internal State Manipulation· AI Security
- New Scheduling Algorithm Cuts LLM Serving Latency by Optimizing Mixed Prefill-Decode Batches· AI
- Researchers teach AI to develop shortcuts through conversation, like humans do· AI
- New AI method trains robots to work smoothly with unfamiliar partners· AI
- How to Measure When AI Writes Like You: A New Framework for Style-Personalized Text· AI
- Extreme compression boosts LLM inference speed up to 7× on CPUs and 6.7× on GPUs· AI
- What makes code examples work in AI prompts? Naming clarity beats formatting· AI
- New dataset aims to make AI image models better at recognizing culture-specific details in low-resource languages· AI
- New Federated Framework Teaches AI Models to Reason Step-by-Step While Keeping Data Private· AI
- How Much Speech Data Do Low-Resource Languages Actually Need? New Study Explores Speech LLM Training Requirements· AI
- Unsupervised neural networks solve CT imaging without ground-truth training data· AI
- Lightweight AI Model Performs Sentiment Analysis with Reasoning on a Budget· AI
- New training method helps AI reason more accurately without hallucinating facts· AI
- New Methods for Privacy-Preserving Project Selection in Community Budgeting· AI
- Comprehensive Survey Maps LLM Safety Risks and Defenses Against Harmful Content· AI Safety
- New benchmark tests whether AI doctors understand medical ethics as well as they know medicine· AI
- How LLMs Feel Emotions Differently: AI Models Treat Feelings as Discrete Categories, Unlike Humans· AI
- New Framework Tackles Equity Gap in Power Restoration After Disasters· AI
- New Test Exposes Hidden Backdoor Vulnerabilities in Federated Learning Systems· AI Security
- New technique shrinks web page data for AI agents without losing capability· AI
- Researchers propose FaST method to personalize AI chatbots with minimal user data· AI
- LLM framework uses spatial reasoning to fix geographic bias in AI predictions· AI
- New Dataset and AI Framework Target Toxic Meme Detection Using Collaborative Tags· AI
- Study reveals which ChatGPT prompting strategies boost productivity in real code generation tasks· AI
- Researchers balance privacy and accuracy in federated learning for cardiac risk diagnosis· AI
- Classical animal learning theory explains how linear transformers attend to information· AI
- RooseBERT brings domain-specific language modeling to political discourse analysis· AI
- LLMs Struggle to Replicate Cross-Cultural Depression Symptom Patterns· AI
- SonicMaster: AI system lets anyone fix audio quality with text instructions· AI
- New scheduling approach cuts LLM latency spikes by up to 77% without moving cached data· AI
- New Framework Adapts Language Models for E-Commerce Click Prediction· AI
- Study Reveals How Artists' Concerns Get Lost in AI Governance Debate· AI
- New Quantization Method Cuts Vision-Language Model Size While Preserving Accuracy· AI
- New method trains AI models to follow physics laws while solving scientific mysteries· AI
- Hybrid AI Model Merges Plant Biology with Deep Learning for Better Crop Forecasting· AI
- New AI Framework Trains Multiple Solver Modules Jointly to Speed Up Complex Optimization Problems· AI
- Marco-Voice: New speech synthesis system blends voice cloning with emotion control· AI
- Reinforcement Learning Optimizes Nearest-Neighbor Search for AI Systems· AI
- New method embeds hidden fingerprints in AI models to prove ownership· AI
- Researchers bypass control flow graphs to speed up pointer analysis· Software Engineering
- Pinterest Boosts Ad Performance Using Graph Neural Networks and Off-Platform Conversion Data· Machine Learning
- New Embedding Model Tailored for Chemistry Searches Dramatically Improves Chemical Literature Retrieval· AI
- Open-Source Solver Implements Cumulative Functions for Complex Scheduling Problems· AI
- New Two-Way Reward Model Improves How AI Systems Evaluate Their Own Reasoning· AI
- Dynamic Multi-Scale Coordination Framework Advances Time Series Forecasting· AI
- New Framework Teaches AI Models to Resolve Contradictory Information Accurately· AI
- New Training Method Improves How AI Diagnoses Abdominal Emergencies with Clinical Evidence Focus· AI
- New Multi-Agent AI System Uses Internal Competition to Improve Trading Decisions· Agentic AI
- Nodes as Autonomous Agents: New Framework Rethinks How Graph Networks Learn· Agentic AI
- Researchers convert multimodal AI models into efficient embedding engines without heavy retraining· Agentic AI
- New Study Tests Whether AI Can Map Poverty Using Satellite Images and Web Data· AI
- New Framework Uses Predictive Models to Stop LLM Agents Before They Act Unsafely· Agentic AI
- Researchers use deep learning and acoustic imaging to detect drones from sound· AI
- Large Language Models Transform Medical Reasoning Through Enhanced Training and Testing Strategies· AI
- New Framework Trains Language Models as Smarter Agents Using Planning-Guided Reinforcement Learning· Agentic AI